Matstat Econometrics Matan - page 32

New comment
 
Доктор #:

For example, even 30% is normal for a trend TS

So it's a bad trend-follower. A good one doesn't lose on a flat.
 
secret #:
So it's a bad trend-following system. A good one doesn't lose on a flat market.

Doesn't drain on the Grail on the Flat. And the Grail is lapis philosophorum.

And for conventional TSthe percentage of profitable deals is not a stand-alone metric, it depends too much on the specifics of TS.

Perhaps you've heard of TS "Time Trade". The percentage of profitable trades tends to 100%. But the creator of this TS is chronically in a state of lawsuits with his investors.

 
secret #:
Well, all systems work this way.
Changing volume is self-defeating, it does not change the average trade if the terms of the trades are the same. An adequate test should only be performed with fixed sizing.

1) It is quite possible to imagine a system on the ideas of fuzzy logic, where in addition to entry direction there is also "certainty" about the correctness of the entry)

2) In classical portfolio theory, when asset prices change, their volumes must change to keep the calculated ratio in money.

3) When trading large enough volumes, sudden changes in positions can be dangerous.

Nevertheless, the percentage of profitable trades is a good metric (in the conditions I described above). In this case it can be considered as a basic one and all the others can be calculated through it. For example, you can find (I hope you know how) a confidence interval for probability of profitable trades)

 
Доктор #:

You may have heard of "Time Trading". The percentage of profitable trades there is close to 100%. Except that the creator of this TS is chronically in court with his investors.

100% is over sitting of course, for normal systems 75-85% is normal.
This is the only bad thing about this metric - it is not good for over sitting)
 
secret #:
100% is over sitting of course, for normal systems 75-85% is the norm.
That's the only thing this metric is bad for - it's not good for overshooting)

rollover is a separate transaction (which is pretty much what it is)

and over-sitting will quickly fall off the pedestal 100%

 
Aleksey Nikolayev #:

1) It is quite possible to imagine a system based on fuzzy logic ideas, where in addition to the direction of entry, there is also a "certainty" that the entry is correct)

Confidence >50% - enter (60, 80 - to taste).
Anything less is not an entry.
It's for pure, uncluttered ej testing)
And in production, yes, you can vary, if religion allows)
 
secret #:
100% is over sitting of course, for normal systems 75-85% is normal.
This is the only thing this metric is bad for - it's not good for overshooting).

The only place where it makes sense is optimization in a small neighborhood of parameters. To compare two different systems is already questionable.

 

The grail is:

1 option. The right entry with a small deposit load and a long hold on the position.

This is investing.

2 option.Correct entry with high deposit load and correct exit with short-term position holding. This is speculative trading with high risk.

But if we divide traders by the level of their preparation into dates from 1 to 10, it appears that the least prepared traders enter the speculative deals.

It's easier to conduct mid-term and long-term trading with low dan. Even Baskakov understood that))

 
Доктор #:

The only place where it makes sense is in a small parameter vicinity. For comparing two different systems it is already questionable.

Why is it questionable? The ege indicator of any system is equity stability, and the simplest indicator of stability is just win%.
 
secret #:
Why is it doubtful? The ejac indicator of any system is equity stability, and the simplest indicator of stability is exactly win%.

The simplest measure ofequity stabilityis RF.The TS can theoretically fail at a rather highwin%.On rare, but highly unprofitable trades. As Alexander_K )))

1...2526272829303132333435363738
New comment